Haar or Haar-like features are formations of rectangles with varying pixel density. Haar features sum up the pixel intensity in the adjacent rectangular regions at specific locations in the detection region. Based on the difference between the sums of pixel intensities across regions, they categorize the different subsections of the image.
Haar-like features have their name attributed to the mathematics term of Haar wavelet, which is a sequence of rescaled square-shaped functions that together form a wavelet family or basis.
Because Haar-like features work on the difference between pixel intensities across regions, they work best with monochrome images. This is also the reason the images used earlier and in also this section are monochrome for better intuition.
These categories can be grouped into three major groups, as follows:
- Two rectangle features
- Three rectangle features
- Four rectangle features
Haar-like Features
With some easy tricks, the computation of varying intensities across the image becomes very efficient and can be processed at a very high rate in real time.